Research on computer-aided diagnosis and implementation of Alzheimer’s disease based on deep learning

MRI and PET are commonly used in the early diagnosis of Alzheimer’s disease (AD). MRI shows the brain's structure, and PET shows how the brain works. A deep learning model is built to diagnose AD. CNN can find small details in 3D brain images. A transformer checks the way that different parts of the brain connect. Thus, computers understand the images better. The database is built by using data from the ADNI database. The data include AD, CN, and MCI. The data were split into training, validation, and testing groups for training and improvement after image processing. The proposed model was evaluated on both two-class and three-class tasks. In the two-class task, the model achieved an accuracy of 94.03% on MRI images and 92.54% on PET images. In the three-class task, the corresponding accuracy was 69.16% for MRI and 68.25% for PET. The results indicate that combining CNN with Transformer improves feature representation for brain images and provides better classification performance than the comparison models. In summary, the proposed CNN-Transformer model efficiently integrates the local structural and global contextual information from MRI and PET images. The proposed method can offer a practical reference for the computer-aided diagnosis of Alzheimer’s disease.